Mean Squared Error (MSE) is the average of the squared differences between predictions and actual values. Squaring the errors penalizes large deviations more than smaller ones. MSE is a common loss function for regression tasks and is differentiable, making it gradient-friendly.